Plasma reactors used to fabricate semiconductor microelectronic circuits can employ RF inductively coupled fields to maintain a plasma formed from a processing gas. Such a plasma is useful in performing etch and deposition processes. Typically, a high frequency RF source power signal is applied to a coil antenna near the reactor chamber ceiling. A semiconductor wafer or workpiece support on a pedestal within the chamber has a bias RF signal applied to it. The power of the signal applied to the coil antenna primarily determines the plasma ion density within the chamber, while the power of the bias signal applied to the wafer determines the ion energy at the wafer surface. One problem with such a coil antenna is that there is a relatively large voltage drop across the coil antenna, which can induce unfavorable effects in the plasma such as arcing. This effect becomes more acute as the frequency of the source power signal applied to the coil antenna is increased, since the reactance of the coil antenna is proportional to frequency. In some reactors, this problem is addressed by limiting the frequency to a low range such as about 2 MHz. Unfortunately, at such lower frequencies, the coupling of RF power to the plasma can be less efficient. It is often easier to achieve a stable high density plasma discharge at frequencies in the range of 10 MHz to 20 MHz. Another disadvantage of operating at the lower frequency range (e.g., 2 MHz) is that the component size of such elements as the impedance match network are much larger and therefore more cumbersome and costly.
Another problem with coil antennas is that efficient inductive coupling to the plasma is generally achieved by increasing the number of turns in the coil which creates a larger magnetic flux density. This increases the inductive reactance of the coil, and, since the circuit resistance (consisting primarily of the plasma resistance) remains constant, the circuit Q (the ratio of the circuit reactance to resistance) increases. This in turn leads to instabilities and difficulties in maintaining an impedance match over varying chamber conditions. Instabilities arise particularly where the coil inductance is sufficiently great so that, in combination with stray capacitance, self-resonance occurs near the frequency of the RF signal applied to the coil. Thus, the inductance of the coil must be limited in order to avoid these latter problems.
These problems have been largely solved by the invention of an inductive coil antenna having multiple interleaved symmetrically arranged conductors spiraling outwardly as set forth in U.S. Pat. No. 5,919,389, filed Jul. 6, 1999 entitled “INDUCTIVELY COUPLED PLASMA REACTOR WITH SYMMETRICAL PARALLEL MULTIPLE COILS HAVING A COMMON RF TERMINAL”, by Xue-Yu Qian et al. By dividing the antenna into multiple conductors in an interleaved symmetric pattern, the voltage drop is reduced because it is divided among plural conductors of the antenna. Thus, the frequency of the source power signal is not restricted as in a conventional coil antenna. This type of coil antenna is referred to in this specification as an “interleaved” coil antenna. Such an interleaved coil antenna is disclosed in various configurations including a flat pancake shape as well as a dome shape or a dome shape with a cylindrical skirt around the side walls or a flat pancake shape with cylindrical skirts around the chamber side wall (U.S. Pat. No. 5,919,389).
One limitation of coil antennas overlying the chamber ceiling (both conventional as well as the interleaved type) is that the mutual inductance between adjacent conductors in the antenna is generally in a horizontal direction generally orthogonal from the vertical direction in which RF power must be inductively coupled to the plasma. This is one important factor that limits the spatial control of the power deposition to the plasma. It is a goal of the present invention to overcome this limitation in the spatial control of the inductive coupling.
Typically with “inner” and “outer” coil antennas, they physically are distributed radially or horizontally (rather than being confined to a discrete radius) so that their radial location is diffused accordingly. This is particularly true of the horizontal “pancake” configuration. Thus, the ability to change the radial distribution of plasma ion distribution by changing the relative apportionment of applied RF power between the inner and outer antennas is limited. This problem is particularly significant in processing semiconductor wafers with larger diameters (e.g., 300 mm). This is because as the wafer size increases, it becomes more difficult to maintain a uniform plasma ion density across the entire wafer surface. The radial distribution of plasma ion density can be readily sculpted by adjusting the radial distribution of the applied magnetic field from the overhead antenna. It is this field which determines plasma ion density. Therefore, as wafer size increases, a greater ability to sculpt or adjust the radial distribution of the applied RF field is required. Accordingly, it would be desirable to enhance the effect of the apportionment of applied RF power between the inner and outer antennas, and in particular to accomplish this by confining each of the inner and outer antennas to discrete or very narrow radial locations.
Another problem encountered with the use of inner and outer coil antennas is that the outer antenna typically has a significantly greater inductance than the inner antenna (because of the longer distances at the outer radii), so that they have vastly different impedances. As a result, the impedances of the two coils are not similar. This problem is more acute as the chamber size increases to accommodate the trend toward larger semiconductor wafers. One way around this problem is to use independent RF power sources to drive the inner and outer antennas. Since each power source has its own impedance match network, a disparity between the impedances of the inner and outer antennas is not a problem. However, another problem arises in that it is difficult or impractical to keep the two independent power sources in phase, so that undesirable effects arise due to the occurrence of constructive and destructive interference between the RF magnetic fields generated by the two antennas as their RF currents wander in and out of phase. This problem is overcome in accordance with one aspect of the invention by employing a novel dual output RF power source having the ability to apportion different RF power levels to its two outputs. However, with such a single RF source, the disparity between the impedances of the inner and outer antennas is again a problem. It would therefore be desirable to facilitate at least near equalization of the impedances of the inner and outer coils without sacrificing the inductive coupling of either.